Total: 1
In the composition process, selecting appropriate single-instrumental music sequences and assigning their track-role is an indispensable task. However, manually determining the track-role for a myriad of music samples can be time-consuming and labor-intensive. This study introduces a deep learning model designed to automatically predict the track-role of single-instrumental music sequences. Our evaluations show a prediction accuracy of 87% in the symbolic domain and 84% in the audio domain. The proposed track-role prediction methods hold promise for future applications in AI music generation and analysis.